Back to News
MOSI•July 9, 2026
MOSS Transcribe-Diarize-0.9B
Open Source
ASR
diarization
multimodal
long-form-audio
### TL;DR
MOSS-Transcribe-Diarize 0.9B is an end-to-end multimodal audio understanding model designed to perform long-form transcription, speaker diarization, and timestamping in a single pass. It eliminates the need for multi-stage pipelines by jointly processing speech to produce structured, speaker-aware text with precise time-aligned segments.
Key Insights & Metrics
Pricing
Free
Cost structure
Version
0.9B
Current release version
Hardware
A CUDA-enabled NVIDIA GPU (such as an RTX 4090 or H100) with sufficient VRAM is required for optimal performance
Compute requirements
Category
Open Source
Licensing model
Region
China
Primary region
Key Features
- End-to-end multi-speaker transcription and diarization
- Real-time or long-form timestamped output with consistent speaker labeling
- Promptable architecture supporting custom instructions and hotwords
- Acoustic event annotation capabilities
- Compatible with vLLM and SGLang for high-performance deployment
Discussion
0
Upvotes
0
Downvotes
0 reviews
Sign in to leave a review
Reviews
No reviews yet. Be the first to review!